The Elasticity of Labor Supply: Evidence from theBiblical Land of Israel
Nadav Ben Zeev∗ Ori Shai†
October 29, 2017
Abstract
The Bible-mandated Sabbath year, during which consumption of agricultural prod-ucts produced in the Biblical land of Israel is prohibited by Jewish law, takes placeevery seventh year and applies only to some part of contemporary Israel. Utilizingan instrumental variable strategy based on the distance of localities from the Biblicalland of Israel, we exploit this old Biblical rule and its temporary, anticipated natureto investigate the elasticity of labor supply. Using the 2008 Israeli Census, we showthat during the Sabbath year income of workers in localities outside of the Biblicalland of Israel is significantly higher than for those inside it. Further, in line with theneoclassical model of labor supply, we find that men’s hours worked are higher forthe former due to their increased income. However, we do not find an effect on la-bor force participation rates. Throughout the analysis we strengthen our findings byalso using difference-in-differences identification strategy; conducting placebo anal-ysis by examining labor market behavior of workers who are less likely to work inagriculture and thus to be affected by the Biblical rule; and showing that there wereno differences in workers’ incomes between localities inside and outside the Biblicalland in the years before or after the Sabbath year.
JEL classification: J22
Key words: Sabbath year, labor supply, transitory income increase
∗Corresponding author: Department of Economics, Ben-Gurion University of the Negev. P.O.B 653,Beer-Sheva 84105, Israel. E-mail: [email protected].†Hebrew University of Jerusalem. E-mail: [email protected] .
1 Introduction
“When you come to the land that I am giving you, the land must be given a rest period, a Sabbath to
God. For six years you may plant your fields, prune your vineyards, and harvest your crops, but the
seventh year is a Sabbath of Sabbaths for the land...”
Book of Leviticus 25
It has been a major concern in the public policy arena whether increasing workers’ wages affects
labor supply. Although many policy makers have tried over the years to increase workers’ labor
supply by setting different rules and applying various reforms, there is still a long lasting debate
in the literature on the effects of increasing workers’ remuneration on labor supply. Contributing
toward resolving this debate on labor supply responsiveness can better our understanding of the
effects of a multitude of economic forces, from labor-income tax policy changes through various
other economic shocks whose propagation mechanism operates partly via labor markets (e.g.,
technology shocks).
In order to identify the elasticity of labor supply (both at the intensive and the extensive mar-
gin), this study takes advantage of an old Biblical rule, called Sabbath year, which prohibits farm-
ers who reside within Biblical Israel from cultivating their land and selling their products every
seven years to the Jewish orthodox population in Israel. Since the borders of Biblical Israel differ
from those of contemporary Israel, farmers who reside outside of Biblical Israel are not affected
by this old rule. Therefore, they are able to sell their products to a larger share of the Israeli
population and to enjoy higher income levels. In other words, we identify the causal effect of a
temporary, anticipated income increase on labor supply by using the income variation between
localities inside and outside Biblical Israel during the Sabbath year.1,2
Importantly, there are two facts that render our research design particularly appealing and
1For more information on the official regulation visit:http://www.moag.gov.il/en/Ministrys%20Units/Foreign%20Trade/Fruit%20and%20Vegetables%20Import%20on%20a%20Sabbatical%20Year%20(Shmita)/Pages/default.aspx.
2More information can be found in the media:https://translate.google.co.il/translate?hl=en&sl=iw&u=https://www.themarker.com/1.2339015&prev=search;http://www.haaretz.com/print-edition/features/fallower-than-thou-1.233480.
1
suitable for addressing our research question. First, the Sabbath year induced wage variation is
transitory and anticipated, thus allowing us to ignore wealth effects and in turn focus on estimat-
ing the substitution effect of wage changes. (Specifically, the labor supply elasticity we are aiming
at identifying in this paper is the Frisch elasticity of labor supply, i.e., the responsiveness of labor
supply to wage changes keeping constant the marginal utility of wealth.3) Second, that farmers
are mostly autonomous in terms of their ability to change their labor supply in response to wage
fluctuations also makes possible and facilitates addressing our research question.
Our main results can be summarized as follows. First, over the course of the 2008 Sabbath
year, incomes among localities outside Biblical Israel increased by 20% relative to localities inside
Biblical Israel. Second, the results reveal that men’s hours worked positively respond to increases
in income from work: a 1% increase in earnings leads to a 0.23%-0.64% increase in hours worked.
However, we do not find any effect on men’s labor force participation rates nor on women’s work-
ing behavior (neither on women’s hours worked nor on women’s participation rates). Throughout
the analysis, we show that individuals did not select to reside outside the borders of Biblical Is-
rael. This robustness check provides further support to the claim that the income change was truly
exogenous, and that the results are not driven by changes in individuals’ behavior just before the
Sabbath year. Moreover, we show that there were no differences in income between localities in-
side and outside Biblical Israel before and after the Sabbath year of 2008, and that there were no
socio-economic and demographic differences between men and women who reside outside and
inside Biblical Israel. These robustness checks demonstrate that the Sabbath year was the true
force behind the results rather than non–Sabbath year related differences in income or generic
differences in socio-economic conditions between the affected and the non-affected workers.
There are two competing theories that try to explain workers’ behavior in response to tran-
3Given that this elasticity is a micro-founded structural parameter, various works have attempted toestimate this elasticity by fitting structural models to both micro data (see, e.g., Heckman and Macurdy(1980), Heckman and MaCurdy (1982), Blundell et al. (1993), Ziliak and Kniesner (2005), French (2005),Kimball and Shapiro (2008), and Blundell et al. (2016)) as well as macro data (see, e.g., Prescott (2002),Ohanian et al. (2008), and Chetty et al. (2011). Below we discuss in detail the literature that has tried to usemore reduced-form, model-free identification approaches to identifying this elasticity, which we view asbeing more comparable to our approach.
2
sitory income changes. Kahneman and Tversky (1979) argue that individuals have reference-
dependent preferences. Therefore, it is possible that workers will choose to work until the reach
a certain level of income. By contrast, in the standard neoclassical inter-temporal model of labor
supply, a transitory positive income change should lead to an increase in labor supply. Hence, in
addition to being crucially important from a policy-making standpoint, pinning down labor sup-
ply elasticity is also vital from an intellectual curiosity standpoint as it can help researchers choose
between these two contrasting, competing theories.
According to the neoclassical inter-temporal model of labor supply, a wage increase induces a
substitution effect by which an economic agent desires to substitute leisure for work, thereby posi-
tivity affecting labor supply. However, the main empirical difficulty in estimating this relationship
is that the wage increase may affect individuals’ wealth by increasing their lifetime earnings (in-
come effect). As such, if non-working activities (e.g., leisure) are normal goods, the wage increase
may have a negative impact on working behavior (along both the intensive and the extensive mar-
gin). Thus, in order to reveal the substitution-induced impact of a wage increase on labor supply,
the income change must be transitory.
In previous years, researchers have tried to address this concern by controlling for the wealth
effect throughout estimations (see, e.g., Altonji (1986)). However, it is still not clear whether these
strategies truly purged the wealth effect of the estimated elasticities. Recently, a few studies have
investigated the effect of a transitory wage increase on labor supply. However, the empirical
results and findings are still mixed. Camerer et al. (1997) and Chou (2002) study the daily labor
supply of New York and Singapore taxicab drivers, respectively, and reveal that cabdrivers tend
to work fewer hours on high-wage days. Farber (2005) finds that daily income has only a small
effect on the decision of taxi cabdrivers to stop working. The author argues that the decision to
stop working is substantially related to the cumulative hours of work to that point.
More recently, Fehr and Goette (2007) conduct a field experiment using bicycle messengers in
Zurich, Switzerland. In their study, the bicycle messengers were free to choose working hours and
effort per hour. The authors find a positive effect of wage increase on hours worked, but a negative
3
effect on effort per hour. The authors argue that workers’ loss aversion contributes significantly
to the negative effort elasticity. Farber (2008) shows that taxi cabdrivers stop working when they
reach their income target level, in accordance with reference-dependent preferences theory. Craw-
ford and Meng (2011) use the datasets from Farber (2005, 2008) and find that reference-dependence
is the main force driving the results in Farbers studies. Doran (2014) finds the New York City taxi
drivers decrease their hours worked in response to a short term wage increase, but do not respond
to a long term wage increase. Stafford (2015) studies the daily labor supply of fishermen in Florida
using the moon phase as an instrument for wage. The author finds that hours worked increase
when earnings are temporary higher. Recently, Farber (2015) uses all trips taken in New York city
taxicabs during the years of 2009 to 2013 and finds that cabdrivers responds positively and in-
crease their hours of work in response to both anticipated and unanticipated increases in income.
Therefore, the author rejects the reference-dependent model. Finally, using non-parametric meth-
ods, Thakral and To (2017) estimate the effect of daily earnings of New York cab drivers on their
labor supply. The authors analyze all cab fares in New York during 2013, and conclude that in
contrast to the neoclassical labor supply model, cab drivers are more likely to stop working when
the cumulative daily earnings are higher.
Most of the aforementioned papers focus on the intensive margin (hours of work). Other
studies identify the effect of income shocks on whether to work at all (the extensive margin). Oet-
tinger (1999) analyzes the labor supply behavior of food and beverage stadium vendors during
one baseball season. Using game attendance as an instrument for wage, the author finds a sub-
stantial positive labor supply elasticity in the 0.55-0.65 range. Goldberg (2016), using data from
rural Malawi, conducts a field experiment and estimates whether unexpected transitory income
shocks affect the probability of accepting employment on particular day. The author reveals that
the labor supply extensive margin elasticity is low (approximately 0.15). Gin et al. (2017) analyze
the extensive margin of labor supply decisions of boat owners in south India. By exploiting ex-
ogenous wage changes (e.g., lunar calendar, variation in internationally determined prices of fish,
and the price of intermediate inputs), the authors find that labor participation is positivity affected
4
by expected earnings. In particular, their estimated labor supply elasticities range between 0.8 and
1.3.
The above literature suffers from four main limitations, which to our knowledge are not wholly
overcome by any one study. The first is that using permanent income changes while controlling
for the wealth effect throughout estimations may not effectively purge the wealth effect of the es-
timated elasticities. The second is that focusing exclusively on the intensive margin is insufficient;
individuals may decide also to join the labor market or to postpone their retirement decision in
response to an income change, rendering it important to also study the impact of income change
on the extensive margin. The third is that conducting controlled experiments to reveal the impact
of a transitory wage increase on labor supply could be problematic given that it is difficult to gen-
erate a realistic labor market environment in an experiment. The last limitation is that some of the
instrumental variables used in literature (e.g., average hourly earnings of other workers) may be
correlated with individuals’ unobserved characteristics that may affect their working behavior.
In this study, we make a step forward in addressing these four issues and contribute to the
literature in four substantial ways. First, we take advantage of an old Biblical rule, one that tar-
gets a substantial part of the Israeli population which has similar characteristics to those of other
cohorts. Second, to the best of our knowledge, this study is the first to exploit a Biblical rule in its
identification strategy. Third, most of the previous studies use data sources that were targeted at
a specific local population (New York cab drivers, food and beverage vendors at a single baseball
stadium, etc.). In this study, the main data source comes from the Israeli 2008 Integrated Census
which provides information on all the localities in Israel and combines data from administrative
sources and from samples collected in surveys.4 Fourth, since our study focuses on a developed
country, the affected group largely has full access to credit markets which in turn enables our
paper to produce finding with valuable implications for labor market policies that can be imple-
mented in other developed countries as well. Moreover, the income change is not associated with
any work efforts, and in contrast to other studies that had short time study periods, the period of
4For more detailed information on the Israeli 2008 Integrated Census visit:http://www.cbs.gov.il/census/census/pnimi page e.html?id topic=7.
5
study in our paper lasts for an entire year.
The remainder of the paper is organized as follows. Section 2 describes the Sabbath year and
the Biblical land of Israel. We introduce the data in section 3. In Section 4 we outline the econo-
metric identification strategy. The results and some robustness checks are discussed in Section 5
and we conclude the analysis in Section 6.
2 The Sabbath Year and the Biblical Land of Israel
According to Jewish tradition, every seventh year the land of Israel must be left alone and get a
rest period for one year during which all agricultural activity (e.g., planting, harvesting, etc.) is
prohibited. Importantly, these restrictions also prohibit the consumption and sale of agriculture
products that were cultivated within the borders of the Biblical land. (Nowadays, the main eco-
nomic implication of the Sabbath year is that it prohibits the religious Jewish population, which
makes up about 20% of Israel’s population, from the consumption of agricultural products grown
inside Biblical Israel.5)
This old rule which is called the Sabbath year (in Hebrew Shmita) is mentioned a few times in
the Bible,6 and it is still observed by ultra-orthodox Jews in the contemporary state of Israel these
days.7 The most important issue of this old, Biblical rule and most relevant to our context lies
in the distinction between Biblical and contemporary Israel. I.e., the Biblical land of Israel differs
in its borders from contemporary Israel, due to which farmers who reside outside the borders of
Biblical Israel do not face any obstacle to selling their agricultural products. Therefore, during the
Sabbath year these farmers experience higher income rates relative to farmers who reside within
the borders of Biblical Israel.
Figure 1 depicts the geographical spread of Israel’s districts. According to the Jewish law,
the HaArava district (which is located below the Dead Sea) is defined as the area outside Biblical
5Source: Israel Central Bureau of Statistics 2007 report entitled ’Characterization of the Jewish Popula-tion by Level of Religiosity Based on Linkage to Educational Institutions’.
6In the books of Exodus, Leviticus, Deuteronomy, etc.7In general we refer to religious Jews as ultra-orthodox Jews.
6
Israel, while all other districts are defined as the area inside Biblical Israel.8,9 Importantly, since the
Sabbath year mainly affects farmers, this paper focuses on Israel’s districts; these districts contain
only rural localities where all the Israeli farmers reside.
The average income in 2008 for districts within and outside Biblical Israel is presented in Figure
2. The figure shows that the average income in the HaArava district (the treated group, depicted
in solid outline) which is located outside Biblical Israel, is higher by approximately 20% relative to
districts that are located within Biblical Israel (the control groups, depicted in dashed outlines). In
sum, the figure reveals that the Sabbath year has a substantial economic impact, whereby localities
that were not affected by the Biblical rule experienced higher incomes compared with the affected
localities.
One should bear in mind that Israel’s size is relatively small.10 Thus, it is less likely that the
treated and the control districts are differently exposed to economic factors such as tourism, re-
cessions, etc. Notwithstanding this reasonable assumption, throughout the paper we use different
ranges of distances from the border of Biblical Israel to assess our findings.
3 Data
In order to identify the effect of a transitory income increase on labor supply, we use the 2008
Israeli Integrated Census that combines data from administrative sources and from samples col-
lected in surveys. The census contains locality-level information on employment, socio-economic
conditions and other demographic characteristics of all the localities in Israel. The 2008 Israeli
Census sampled 20% of the Israeli population (which was approximately 7.3 million people in
2008) through personal interviews combined with administrative data from the population reg-
istry. Since localities with a population size of less than 300 are fully sampled, more than 40%
8HaArava district is also called Arava in some publications.9For more information on the official regulation visit:
http://www.moag.gov.il/en/Ministrys%20Units/Foreign%20Trade/Fruit%20and%20Vegetables%20Import%20on%20a%20Sabbatical%20Year%20(Shmita)/Pages/default.aspx.
10According to the CIA’s fact book, Israel’s size is slightly larger than New Jersey, which is the fourthsmallest state in the U.S.
7
of the HaArava district (our treated group, which contains 7 relatively small localities) was sam-
pled.11 The Israeli population census was conducted in 2008, and since the penultimate Sabbath
year was also observed in 2008, it provides us with a unique opportunity to identify the impact of
a transitory income increase on labor supply.
As previously mentioned, during the Sabbath year, a substantial part of the Israeli population
does not purchase agricultural products that were cultivated within Biblical Israel. Since Israeli
farmers reside within rural areas, we focus our analysis on all the rural localities in Israel. Each
rural locality/community contains approximately 850 people who live in individual farms, which
are called in Hebrew Kibbutzim or Moshavim. Note that during the 1990’s most of the Kibbutzim
became private and each one of their members earns his own income and pays his own costs.12,13
Importantly, farmers who are self-employed can allocate their time between working and
leisure and decide whether to continue working or to retire. During the Sabbath year of 2008,
third of the workers in the HaArave district, were self-employed. Since these workers can deter-
mine their labor supply in response to a transitory income increase, the effect of the Sabbath year
is potentially prominent.
The 2008 Israeli census contains data on hours worked and labor force participation rates (de-
fined separately for men and women for every rural locality in Israel), but it does not have infor-
mation on incomes. However, since each locality is located within a specific district, we merge
each locality with district-level administrative information on income and other socio-economic
and demographic characteristics taken from the Central Bureau of Statistics (CBS) Local Authori-
ties in Israel dataset.14 See Appendix A for more details on the variables used in our analysis.
Table 1 displays summary statistics for the 2008 Israeli Census. Panel A of Table 1 shows that
the average income in rural districts in Israel is NIS 7,538. On average, in each rural locality 25
11For more information on the Israeli 2008 census visit:http://www.cbs.gov.il/census/census/main mifkad08 e.html .
12See Ebenstein et al. (2015) for more information on the process of privatization in the Kibbutzim.13Each member of the Moshavim (and nowadays in most of the Kibbutzim) is a private entity. Namely,
each member is responsible for his own costs and expenses.14Our data on districts’ income is taken from here: http://www.cbs.gov.il/reader/?MIval=cw usr view
SHTML&ID=446.
8
percent of all people hold an academic degree. The average men working week is 46 hours, while
women average workweek is only 35 hours. Men’s labor force participation rate is 73 percent,
whereas women’s labor force participation rate is only 68 percent. There is a mean distance of 229
kilometers separating rural localities from the borders of Biblical Israel, and 59 percent of the rural
localities’ population is between the age of 18 and 64.
Panels B-D display similar results using localities with smaller distance ranges from the border
of Biblical Israel. As can be seen, the locality-level characteristics remain similar when we limit
the distance from the border of the Biblical Israel.15 In particular, the main key variables: average
income, men’s employment rates and men’s working week, remain almost the same when using
different borders from the HaArava district, suggesting that we can use localities that are located
within various ranges from the borders of Biblical Israel as control groups in the upcoming anal-
ysis. (Part of the agriculture workers in Israel are foreign; therefore, they are not included in our
datasets.)
Finally, it is important to stress that we also examine labor force participation rates (in addition
to hours worked) because farmers can decide whether to continue working, re-enter the labor
market or to retire. Thus, is it also important to examine labor force participation in response to a
transitory income increase. Notably, in the following sections we will separate the analysis to men
and women, since 81% of the agriculture labor market in Israel is composed by men.16
4 Identification Strategy
We aim to estimate the effect of a transitory income increase on labor supply. Therefore, our basic
econometric model is:
LSj = α0 + α1 Incomes + α2Xj + uj, (1)
15We use the 280 kilometer as the smallest range from the border of the Biblical Israel because there aresome Bedouins localities within smaller ranges which may differ in their characteristics from the HaAravalocalities.
16This statistic appears in Figure E-4 of this CBS document:http://www.cbs.gov.il/www/statistical/mw2013 e.pdf.
9
Where LSj represents either average hours worked or labor force participation rates in locality
j separated for men and women; Incomes is the average income in district s; Xj is a vector of
locality-level control variables that includes: establishment year, the proportion of people that
holds an academic degree and the proportion of people between the age of 18 and 64 in locality j
(presented in Table 1); and uj represents unobservable characteristics of locality j.
One could proceed by estimating Equation (1) via OLS. The possible drawback of such an
approach is that locality labor supply probably affects average locality income. Moreover, income
might be correlated with unobserved locality characteristics (e.g., economic conditions) that also
affect labor supply. Therefore, OLS estimate of α1 would likely to be inconsistent.
We address this concern by exploiting the old Biblical rule (Sabbath year) as an instrument
variable. Toward this end, we define the border of Biblical Israel as the threshold and construct an
instrument variable for income as follows:
Z =
1, if locality j is located outside the Biblical land,
0, otherwise.(2)
I.e., the instrument is an indicator variable that receives the value 1 if locality j is located
outside Biblical Israel (in the HaArava district) and zero otherwise. Importantly, throughout all
estimations we control for the distance of the localities from the borders of Biblical Israel. Thus,
our econometric method can be considered as a regression discontinuity design.17,18 Controlling
for the distance from the borders of Biblical Israel accounts for the differences between the treated
and the control localities (e.g., weather, pollution, labor market characteristics, etc.). Moreover,
throughout estimations we provide several robustness checks by limiting the distance from the
border of Biblical Israel.19
17See Lee and Lemieux (2010) for more information.18I.e., the distance from the borders of Biblical Israel is the running variable in our econometric specifi-
cation.19Importantly, one may argue that our instrumental variable identification strategy should control for
different effects on both sides of the cutoff (the border of Biblical Israel). However, there is no reason whylocalities in the HaArava district would be affected differently by the Sabbath year. The impact variation ofthe Sabbath year appears only in localities inside the Biblical land, therefore, throughout all estimations we
10
The identifying assumption is that there is no other factor that affected the treated group’s (lo-
calities in the HaArava) working behavior in 2008 other than the income change. We will support
this claim by showing below that there were no other discrete changes rather than the income
change. Moreover, previous studies acknowledged the concern of ”division bias”. However, us-
ing the borders of Biblical Israel as an instrumental variable addresses this issue.20
One should notice that the unit of measure in our analysis is the locality-level. The focus on the
locality-level is necessary, since our identification strategies rely on the distance of each locality
from the border of the Biblical Israel. Thus, we are interested in placing similar emphasis on each
locality rather than to be affected by the number of observations in large localities. Moreover, the
focus of the analysis on measures at the locality-level is in line with some of the recent literature
(e.g., Goldberg (2016)).
We end this section with an important comment on the interpretation of our estimated labor
supply responses. As previously explained, throughout the analysis, we use localities within the
Biblical land as the control group. These localities were affected indirectly by the Sabbath year
and experienced a decrease in their income relative to the income increase outside the Biblical
land (since many ultra orthodox Jews avoid purchasing agricultural products that were cultivated
inside the Biblical land). As such, our identification strategy captures the entire effect of the income
increase in localities outside the Biblical land on hours of work relative to localities inside it.
5 Results
5.1 First Stage Results
We aim to identify the causal effect of a transitory income increase on men’s labor supply by ex-
ploiting the old Biblical rule as an instrument variable for income. For this econometric approach
to succeed, a strong correlation must exist between the instrument described above and income,
control the distance from the HaArava district using the KM variable.20See the discussion in Farber (2005).
11
the endogenous explanatory variable. Table 2 demonstrates that this is indeed the case, with F-
statistics in the First Stage regressions exceeding 10 (Staiger and Stock (1997)).
These findings justify the use of this instrument to estimate the causal effect of income on labor
supply. Moreover, as can be seen in Panels A through D of Table 2, our results do not change when
we limit the sample size to localities within smaller distance ranges from the borders of Biblical
Israel. The robustness of the results to the use of different distance ranges, and to the inclusion or
exclusion of several control variables shows that our main findings are not due to the treatment
variable being correlated with observable factors that affect income.
5.2 Intensive Margin
OLS versus 2SLS Results. Table 3 presents the estimated effects for Equation (1) from OLS
and 2SLS for men. In this analysis we compare the treated group (localities outside Biblical Israel)
with localities inside Biblical Israel within different distance ranges from the borders of Biblical
Israel, while controlling for the distance from the borders in the regressions. As can be seen in
the table, the OLS coefficients are positive and range between 0.03 and 0.06, but only one of them
is statistically significant at acceptable levels. Importantly, the OLS estimates do not take into
account the fact that individuals’ hours worked may affect their earnings, or that other factors
may affect individual’s hours worked and also their earnings (i.e., the endogeneity problem).
The 2SLS analysis addresses this concern and reveals that higher income increases men’s hours
worked as indicated by the statistically significant 2SLS coefficients. Specifically, the 2SLS coeffi-
cients range between 0.23 and 0.64, suggesting that a 1% increase in earnings leads to a 0.23%-
0.64% increase in hours worked. In addition, that the coefficients are similar across estimates
reinforces the validity of using the borders of Biblical Israel during the Sabbath year as an instru-
ment variable for earnings. Moreover, the robustness of the results to the use of different distance
ranges shows that our main findings are not due to the control group that was used. (Importantly,
although not shown, similar results were obtained when other distance ranges from the borders
of Biblical Israel were used, including ranges that are smaller than the 280 km range.)
12
One may argue that since hours worked decrease with age, we should not expect to find an
impact of income on hours worked among localities with a considerable elderly population share.
Therefore, we divide the sample into two subcategories. Panel A of Table 4 shows that in localities
where the percentage of men aged 18 to 64 is higher than the first quartile, the 2SLS estimate is
statistically significant and its sign and magnitude are large. By contrast, as can be seen in Panel
B of Table 4, in localities where the percentage of men aged 18 to 64 is lower than the first quar-
tile the 2SLS estimate is statistically insignificant and negative, suggesting that income does not
impact men’s hours worked among relatively old populations. These findings reveal that income
positively affects hours worked among relatively large working age populations. As expected,
labor supply does not respond when we restrict the sample to localities where income plays a
smaller role in changing hours worked. These results reinforce our previous findings that the Sab-
bath year affects hours worked through the income increase rather than a spurious difference in
overall hours worked.
It is possible that the effect of the transitory income increase on hours worked would be more
prominent among the less wealthy individuals. Since we lack data on financial assets, we use the
proportion of people who own a home as a proxy for locality’s wealth. Panel C of Table 4 reveals
that the 2SLS estimate is statistically significant and its sign and magnitude are large in localities
where the proportion of people who own a home is lower than the first quartile. By contrast,
as can be found in Panel D of Table 4, the 2SLS coefficient for wealthy localities is statistically
insignificant, indicating that the transitory income increase does not affect hours worked of rela-
tively wealthy men. As we would expect, these results suggest that the transitory income increase
is more likely to impact men whose paid work plays a substantial part of their earnings.
As previously explained, throughout estimations we control for the linear distance term (KM),
which represents distance from the border of Biblical Israel. However, following Gelman and
Imbens (2014), one may argue that we should control for second order polynomials as well. Panel
E of Table 4 presents the estimated effects of Equation (1) when adding a quadratic term for the
distance variable. As can be seen, the 2SLS results remain the same, i.e., there is a positive effect of
13
an income increase on hours worked. The robustness of the results to the inclusion of high order
polynomials indicates that our main findings are not due to the chosen econometric specification.
One may raise the concern that the structure of the labor market differs between localities and
that this in turn may bias our results. To alleviate this concern, we examine whether the results
remain the same when controlling for the proportion of people who work in the industry sector.
Panel A of Table B.1 reveals that the 2SLS estimates remain the same when controlling for labor
market characteristics. These results suggest that our findings do not stem from differences in
labor market conditions.
Alternative Identification Strategy: Difference-in-Differences. Our previous analysis and
findings rest on the assumption that, after controlling for the distance from the borders of Biblical
Israel, localities outside and inside Biblical Israel are very similar to one another. However, if this
assumption does not hold, our previous findings are spurious.
Therefore, we employ a difference-in-differences identification strategy, in which we compare
labor supply behavior (at the intensive and extensive margin) between men who reside inside and
outside the HaArava district. As previously mentioned, during the Sabbath year, the HaArava
district experiences a wage increase relative to all other districts because HaArava farmers are not
prohibited from selling their products to ultra-religious consumers. Since 81% of the agriculture
labor market in Israel is composed by men (see Footnote 16), we expect that men who reside
outside Biblical Israel change their working behavior during the Sabbath year relative to men
inside the Biblical land (because their income is relatively higher). In this specification, men who
reside in the HaArava served as the ”after treated”, while men who reside outside the HaArava as
the ”before treated”. By contrast, women are only 19% of the people who work in the agriculture
sector and are also less likely to be self-employed;21 therefore, women who reside outside and
inside Biblical Israel should not be affected by the Sabbath year and are thus served as the control
21Only 30% of the self-employed are women, which in turn makes them less likely tochange their labor supply in response to an income shock. For more information, please see:http://adva.org/wp-content/uploads/2015/02/%D7%A2%D7%A6%D7%9E%D7%90%D7%99%D7%9D-%D7%90%D7%A0%D7%92%D7%9C%D7%99%D7%AA-11.pdf.
14
group.
In other words, using women as a control group would capture any difference in labor market
conditions and habits between localities inside and outside the Biblical Israel (e.g., labor market
regulations, working conditions, etc.). This identification strategy would be valid only if men
and women who reside outside the Biblical land are very similar to men and women who live
inside the Biblical land (respectively) along various non-income related dimensions. Panel A of
Table 5 demonstrates that this is indeed the case. The various socio-economic control variables
(the proportion of men and women who hold an academic degree, men and women median age,
etc.) are very similar for men who reside inside and outside Biblical Israel. Similar results hold
for women. Thus, Panel A of Table 5 indicates that we can compare between men (women) who
reside inside the HaArava to men (women) who reside outside the HaArava.
Specifically, the goal in this part of the analysis is to compare the difference between hours
worked of men who reside outside and inside Biblical Israel with the corresponding difference in
hours worked of a control group that is less likely to be affected by the Sabbath year (e.g., women).
Thus, we employ a difference-in-differences identification strategy as follows:
LSj,m = α1 + menm + HaAravaj + α2DiDj,m + α3Xj,m + vj,m, (3)
Where LSj,m measures the labor supply for men or women in locality j (hours worked or participa-
tion rates); HaAravaj is an indicator variable for whether locality j is in the HaArava district, i.e.,
this variable accounts for the differences between localities inside and outside the HaArava dis-
trict; and DiDj,m is an indicator variable for whether group m are men who reside in the HaArava
district. Throughout estimations we control for men fixed effects (menm) which accounts for the
differences between men and women. The Xj,m captures locality-specific characteristics such as:
the proportion of people between aged 18 and 64, the proportion of people who hold an academic
degree and establishment year (presented in Table 1). The vj,m represents the error term. Our main
interest is in the α2 coefficient which captures the causal effect of a transitory income increase on
labor supply as a result of the Sabbath year.
15
Panel B of Table 5 displays the difference-in-differences results using Equation (3). As pre-
viously explained, the difference-in-differences analysis compares the difference between hours
worked of men who reside outside and inside Biblical Israel and the corresponding difference for
a control group that is less likely to be affected by the Sabbath year (i.e., women in all the locali-
ties). According to the difference-in-differences estimates, the difference between hours worked of
men who reside outside and inside Biblical Israel with the corresponding difference in women’s
hours worked is significantly positive during the Sabbath year. I.e., the Sabbath year generates
a significant increase in hours worked of men residing outside Biblical Israel relative to all other
men and women. The results remain the same when we add different control variables and limit
the range from the border of the Biblical Israel. The robustness of the results to the use of different
distance ranges and to the inclusion of several control variables shows that our main findings are
not due to the treatment variable being correlated with other factors that affect working behavior.
5.3 Extensive Margin
We now turn to estimating the effect of a transitory income increase on labor force participation
rates using our baseline 2SLS identification approach and the difference-in-differences identifica-
tion strategy.
OLS versus 2SLS Results. To estimate the effect of a transitory income increase on labor force
participation rates, we use Equation (1) above and replace the hours worked outcome variable
with labor force participation rates from the 2008 Israeli Census. As can be seen in columns (2)
and (4) of Table 6, the 2SLS coefficients are statistically significant and positive suggesting that
participation rates are positively associated with an income increase. However, using smaller
distance ranges from Biblical Israel (as can be seen in Columns (6) and (8) of Table 6) results in
the 2SLS coefficients sharply decreasing and becoming statistically insignificant, suggesting that a
transitory income increase does not affect labor supply on the extensive margin. By and large, it
is apparent that the results from Table 6 are not conclusively consistent with the hypothesis that a
16
transitory income increase affects participation rates.
Difference-in-Differences. We conduct a similar analysis to that applied to hours worked but
we use instead labor force participation rates as our outcome variable. As previously explained,
we compare participation rates between men who reside outside and inside Biblical Israel, against
a control group that is less likely to be affected by the Sabbath year (e.g., women in all the lo-
calities). Panel C of Table 5 presents the difference-in-differences estimates and reveals that the
Sabbath year does not affect labor force participation, as indicated by the statistically insignificant
difference-in-differences coefficients. I.e., a transitory income increase does not affect labor supply
on the extensive margin.
The results remain the same when we use different distance ranges from the border of Biblical
Israel and control for different confounding factors throughout estimations, reinforcing our find-
ings that there is no substantial impact of an income change on participation rates among men.
Overall, taken together, the results from the 2SLS and difference-in-differences analysis are consis-
tent with the hypothesis that a transitory income increase does not affect labor force participation
rates.
5.4 Robustness Checks
Placebo Tests. Up to now, we have assumed that there were no additional factors that affected
the treated group’s (the HaArava district) labor supply in 2008 other than the income change.
However, one may wonder whether the increase in men’s workweek stems from the effect of the
Sabbath year income change, or from differences in unobservable characteristics between localities
inside and outside the Biblical land. In order to address this concern, we estimate the effect of an
income change on women’s labor supply (i.e., hours worked and participation rates) using the
borders of Biblical Israel as an instrument for income. The assumption underlying this estimation
is that the Sabbath year income change would not affect women’s labor supply since women are
less likely to work in the agriculture sector.
17
Table 7 shows that this is indeed the case, as indicated by the statistically insignificant 2SLS
coefficients. Using different distance ranges from the border of Biblical Israel, women’s labor
supply (both at the extensive and the intensive margin) does not respond to the transitory income
increase caused by the Sabbath year, reinforcing our previous finding that the Sabbath year was
the true force behind the results rather than spurious differences between localities. Notably, the
insignificant impact of the transitory income change on women’s labor supply supports the 2SLS
identification assumption that there were no other discrete changes at the border of Biblical Israel
(at the cutoff) rather than the income change.
Anticipation Effects. One may also argue that since the Sabbath year is observed every sev-
enth year, individuals would anticipate the upcoming income increase and reside outside the bor-
der of Biblical Israel. Therefore, we examine whether the proportion of people who lived within
the locality five years ago has increased in the treated group relative to the control groups (lo-
calities inside Biblical Israel, within different distance ranges from the Biblical border). I.e., we
estimate the following model:
Lived within Localityj = ρ0 + ρ1Distances + ρ2HaAravaj + $j, (4)
where Lived within Localityj is the proportion of people who lived in locality j five years ago;
Distances measures the distance of district s from the borders of Biblical Israel; and HaAravaj is
an indicator variable that receives the value 1 if locality j is located outside Biblical Israel and zero
otherwise. Our main coefficient of interest is ρ2 which measures the effect of being outside Biblical
Israel on the proportion of people who lived in locality j five years ago.
Column (1) of Table 8 reveals that the proportion of people who lived within the locality five
years ago is similar in both the treated and the control groups. Similar analysis was conducted
using as the outcome variable the proportion of people who lived outside the locality five years
ago. As can be seen in column 2 of Table 8, similar proportions of people who lived outside the
locality five years ago were found among the treated and the control groups as indicated by the
18
insignificant results of Equation (4). These findings suggest that individuals did not systematically
opt to reside in localities outside Biblical Israel in anticipation of the income increase there.
In our instrumental variable design, the identifying assumption is that there are no other dis-
crete changes that occur at the cutoff (at the border of the Biblical land), other than the income
change. We provide further support to this claim by estimating the effect of being outside the Bib-
lical land on education levels using Equation (4) above. Column 3 of Table 8 shows that there are
no differences in education levels between the treated and the control groups. As can be seen in
panels A to D, similar results were obtained when using localities within different distance ranges
from the Biblical border. Moreover, Figure 3 contrasts the percentage who hold an academic de-
gree in localities inside and outside the Biblical land. As can be seen in the figure, both the treated
and the control districts had similar education levels in 2008. These findings demonstrate that
there were no other discrete changes that occur at the cutoff other than the income change, which
reinforce our previous conclusions that the Sabbath year affected hours worked only through the
transitory income increase.
We also estimate the effect of living outside the Biblical land on the proportion of men who
hold an academic degree using Equation (4) above. This is a similar estimation exercise to that
used in the third column of Table 8 only that it focuses only on men’s education, disregarding
women’s education. This additional estimation exercise is potentially important given this paper’s
focus on men’s labor supply. As can be seen in Panel B of Table B.1, there are no differences in
the proportions of men who hold an academic degree between localities inside and outside the
Biblical land indicating a similarity between the treated and the controlled localities along this
important dimension.
We now put forward an important argument which can ameliorate the general concern that
anticipation of the Sabbath year may lead some people to change their behavior (e.g., postpone
retirement) prior to the Sabbath year. The argument goes as follows: Although the Sabbath year
timing is perfectly predicted, its financial impact is completely uncertain. Therefore, individuals
are less likely to change their working behavior prior to the Sabbath year. E.g., in past Sabbatical
19
years, Israel consumed products from the Gaza strip. However, since the Hamas terror organiza-
tion took over the Gaza strip in 2007, Israel decreased its imports from the Gaza strip. Thus, the
financial impact of the Sabbatical year is not fully anticipated in advance.22
Pre- and Post-Treatment Differences. Throughout the analysis, we argue that the Sabbath
year increased earnings of the treated district (HaArava). However, if the income of the affected
district in the years preceding or following the Sabbath year is significantly different from the
control districts, our identification strategies will yield invalid estimates. Panels A and B of Table
9 show that there are no significant differences in incomes between the HaArava district and all
other districts in the years before (2002-2007) and after (2009-2013) the Sabbath year of 2008, indi-
cating that workers inside and outside Biblical Israel earned almost the same in the years before
and after the Sabbath year of 2008. Moreover, we expect that during the Sabbath year the affected
district would experience a larger income increase relative to other districts and to other years.
Therefore, we estimate the following equation:
Incomes,t = φ0 + φ1KMs + φ2Yeart + φ3DiDs,t, (5)
where Incomes,t is the average income in district s at time t;23 KMs measures the distance of district
s from the borders of Biblical Israel; Yeart are year fixed effects; and DiDs,t is an indicator variable
that receives the value 1 if district s is located outside Biblical Israel in the year of 2008 and zero
otherwise. Our main coefficient of interest is φ3 which measures the effect of being outside Biblical
Israel in 2008 on income. (In other words, we use a difference-in-differences approach.)
Panel C of Table 9 shows that the DiDs,t coefficient is positive and statistically significant,
indicating that during the Sabbath year the affected district experienced a sharp increase in its
earnings relative to other districts and to other years. Figure 4 contrasts the average income of the
treated district with those of all other districts in the years before (2002-2007), during (2008) and
22See, e.g.: http://www.jpost.com/Israel/IDF-to-import-Gaza-produce-on-shmita.23We converted incomes into 2011 terms using the following formula: prices in 2011 terms = prices in
current year×(CPI in 2011)/(CPI in current year).
20
after (2009-2013) the Sabbath year. Again, the figure reveals a sharp difference in average income
between the treated and the control districts in 2008 (the difference is statistically significant). In
contrast, there were no statistical significant differences in income between the treated and the
control districts in the years before or after the Sabbath year. These findings indicate that the
increase in income occurs only in 2008, and that there were no differences in incomes between the
HaArava and other districts before or after the Sabbath year, demonstrating that the Sabbath year
of 2008 was the true force behind the results rather than locality-specific differences in income.
Unfortunately, we do not have district/locality-level data on hours worked in the years before
and after 2008 and are thus unable to estimate a similar specification to Equation (5) for men’s
hours worked. Notwithstanding this limitation, we find it unlikely that our results are driven
by generic, non-income related differences between men’s hours worked inside and outside the
HaArava for the following two reasons. First, the difference between men’s hours worked inside
and outside the HaArava is found to be significantly positive also in comparison to the corre-
sponding difference for women’s hours worked. This lends credence to the view that our main re-
sults are not driven by potential non-income related differences between labor supply in localities
within and outside Biblical Israel. Second, income seems to be the only considered demographic
variable to vary between the treatment and control groups in the Sabbath year. This second re-
sult compliments the first in enhancing our confidence in the interpretation of our results as being
driven mainly by income differences only during the Sabbath year between HaArava and non-
HaArava based localities.
6 Conclusion
This paper has shown that temporary wage increases lead to a rise in men’s hours of work while
having an insignificant effect on labor force participation rates. While the former result is consis-
tent with the neoclassical theory of labor supply rather than the reference-dependent preferences
based theory of labor supply, the latter result emphasizes the need to construct models that are
21
capable of producing implications consistent with both results. To obtain these results, we em-
ployed two different identification strategies, instrumental variable and difference-in-differences,
and exploited a unique research design based on the Sabbath year which dictates that every sev-
enth year consumption of agricultural products produced in Biblical Israel is prohibited. We have
shown that during the Sabbath year, income among localities outside Biblical Israel was signif-
icantly higher relative to localities inside it, and that there were no income differences between
localities during the years before or after the Sabbath year.
Throughout the analysis we reinforce the causal interpretation of our main findings by show-
ing that anticipation did not lead individuals to reside outside the borders of Biblical Israel nor
to change their behavior prior to the treatment. Moreover, we did not find any differences in
women’s hours worked (as well as other, additional demographic variables commonly focused
upon by the literature) between the treated and the control localities, suggesting that there were
no other discrete changes at the cutoff (at the border of Biblical Israel) during the Sabbath year
apart from the income change.
The debate in the literature regarding the casual effect of income on labor supply still exists.
This study sheds some light on this relationship and shows that income affects labor supply only
on the intensive margin. Finding the determinants of the mixed results and why various workers
studied in the literature respond differently to an income increase, however, is left for future work.
The policy implications of our results are potentially valuable. They suggest that putting to-
gether wage-increasing policy initiatives (e.g., through tax reforms) can be effective in terms of
raising labor supply. But, importantly, their benefits are likely to derive from the response of
hours worked rather than labor force participation rates. However, our analysis does carry with it
an important caveat which applies to many labor supply studies: caution need be taken when at-
tempting to generalize this paper’s results. The external validity of our empirical analysis is likely
limited given our particular focus on farmers’ labor supply behavior. (And our focus on farmers
has naturally led to our focusing only on men’s hours worked, preventing us from also studying
the labor supply behavior of women.) Notwithstanding this caveat, we are still confident that our
22
results contribute to our understanding of labor supply behavior.
23
Appendix A Data
A.1 Variables Definitions.
Average hours worked. Locality-level weekly average of hours worked from the 2008 Israeli
Census.
Labor Force Participation Rates. Locality-level share of population (those aged 15 and over,
not including single residents in institutions) who worked during 2008, in any type of work, or
did not work at all but actively sought work. The source of this data is the 2008 Israeli Census.
Income. District-level income was calculated by multiplying the average employee pay and
the average self-employment pay by their relative sizes (by the number of employees and self-
employed people, respectively). The source of this data is district-level administrative informa-
tion on income and other socio-economic and demographic characteristics taken from the Central
Bureau of Statistics (CBS) Local Authorities in Israel dataset.
Distance (KM). Distance in kilometers for each locality from the Tzofar locality in the center
of the HaArava district.24
Wealth. The wealth variable is defined as the proportion of people who own a home as a proxy
for locality’s wealth. The source of this data is the 2008 Israeli Census.
Other Controls. Education level: the proportion of people who hold an academic degree;
Working-age population: the proportion of people between aged 18 and 64; Establishment year:
the year the locality was founded. The source of these data series is the 2008 Israeli Census.
24We calculate the distance using this website: http://www.tremp.co.il/distance/distance.php.
24
Appendix B Additional Robustness Checks
Table B.1: Men’s Education as Outcome Variable and Controlling in the Baseline Regres-sion for the Proportion of People who Work in the Industry Sector.
Panel A: Mens LogWeekly hours worked
(1) (2) (3) (4)Entire Sample 320 km 300 km 280 km
Log Income 0.800** 0.340*** 0.377*** 0.608***(0.392) (0.0614) (0.0795) (0.125)
Observations 587 478 455 383
Panel B: DependentVariable - Percentage ofMen with an Academic
Degree(1) (2) (3) (4)Entire Sample 320 km 300 km 280 km
Indicator for beingOutside Biblical Israel
-4.409 0.576 -3.383 -4.251
(3.544) (3.803) (3.758) (4.046)
Observations 819 682 654 558
Notes: Panel A of this table presents OLS and 2SLS results for each sample. All the regressionsinclude the distance in kilometers from the HaArava district and other control variables, whichare: the proportion of people aged 18 to 64; the proportion of people who hold academic degree;year of establishment; and the proportion of people who work in the industry sector. Robuststandard errors are in parentheses. Panel B of this table presents the effect of being outside BiblicalIsrael on men’s education for different distances (entire sample, within 320 km, 300 km, and 280km) from the HaArava district. All the regressions include the distance in kilometers from theHaArava district. * indicates significance at the 10% level; ** indicates significance at the 5% level;and *** indicates significance at the 1% level.
25
References
Altonji, J. G.: 1986, Intertemporal Substitution in Labor Supply: Evidence from Micro Data, Journal
of Political Economy 94(3), 176–215.
Blundell, R., Meghir, C. and Neves, P.: 1993, Labour supply and intertemporal substitution, Journal
of Econometrics 59(1), 137 – 160.
Blundell, R., Pistaferri, L. and Saporta-Eksten, I.: 2016, Consumption inequality and family labor
supply, American Economic Review 106(2), 387–435.
Camerer, C., Babcock, L., Loewenstein, G. and Thaler, R.: 1997, Labor supply of new york city
cabdrivers: One day at a time, The Quarterly Journal of Economics 112(2), 407–441.
Chetty, R., Guren, A., Manoli, D. and Weber, A.: 2011, Are micro and macro labor supply elastici-
ties consistent? a review of evidence on the intensive and extensive margins, American Economic
Review 101(3), 471–75.
Chou, Y. K.: 2002, Testing alternative models of labour supply: Evidence from taxi drivers in
singapore, The Singapore Economic Review 47(01), 17–47.
Crawford, V. P. and Meng, J.: 2011, New york city cab drivers’ labor supply revisited: Reference-
dependent preferences with rational-expectations targets for hours and income, American Eco-
nomic Review 101(5), 1912–32.
Doran, K.: 2014, Are long-term wage elasticities of labor supply more negative than short-term
ones?, Economics Letters 122(2), 208 – 210.
Ebenstein, A., Hazan, M. and Simhon, A.: 2015, Changing the cost of children and fertility: Evi-
dence from the israeli kibbutz, The Economic Journal .
Farber, H. S.: 2005, Is tomorrow another day? the labor supply of new york city cabdrivers, Journal
of Political Economy 113(1), 46–82.
Farber, H. S.: 2008, Reference-dependent preferences and labor supply: The case of new york city
taxi drivers, American Economic Review 98(3), 1069–82.
Farber, H. S.: 2015, Why you cant find a taxi in the rain and other labor supply lessons from cab
drivers, The Quarterly Journal of Economics 130(4), 1975–2026.
26
Fehr, E. and Goette, L.: 2007, Do workers work more if wages are high? evidence from a random-
ized field experiment, American Economic Review 97(1), 298–317.
French, E.: 2005, The effects of health, wealth, and wages on labour supply and retirement be-
haviour, The Review of Economic Studies 72(2), 395–427.
Gelman, A. and Imbens, G.: 2014, Why high-order polynomials should not be used in regression
discontinuity designs, Working Paper 20405, National Bureau of Economic Research.
Gin, X., Martinez-Bravo, M. and Vidal-Fernndez, M.: 2017, Are labor supply decisions consistent
with neoclassical preferences? evidence from indian boat owners, Journal of Economic Behavior
& Organization 142(Supplement C), 331 – 347.
Goldberg, J.: 2016, Kwacha gonna do? experimental evidence about labor supply in rural malawi,
American Economic Journal: Applied Economics 8(1), 129–49.
Heckman, J. J. and MaCurdy, T.: 1982, Corrigendum on a life cycle model of female labour supply,
The Review of Economic Studies 49(4), 659–660.
Heckman, J. J. and Macurdy, T. E.: 1980, A life cycle model of female labour supply, The Review of
Economic Studies 47(1), 47–74.
Kahneman, D. and Tversky, A.: 1979, Prospect theory: An analysis of decision under risk, Econo-
metrica 47(2), 263–291.
Kimball, M. S. and Shapiro, M. D.: 2008, Labor supply: Are the income and substitution effects
both large or both small?, Working Paper 14208, National Bureau of Economic Research.
Lee, D. S. and Lemieux, T.: 2010, Regression Discontinuity Designs in Economics, Journal of Eco-
nomic Literature 48(2), 281–355.
Oettinger, G. S.: 1999, An empirical analysis of the daily labor supply of stadium venors, Journal
of Political Economy 107(2), 360–392.
Ohanian, L., Raffo, A. and Rogerson, R.: 2008, Long-term changes in labor supply and taxes:
Evidence from oecd countries, 1956–2004, Journal of Monetary Economics 55(8), 1353–1362.
Prescott, E. C.: 2002, Prosperity and depression, American Economic Review 92(2), 1–15.
Stafford, T. M.: 2015, What do fishermen tell us that taxi drivers do not? an empirical investigation
of labor supply, Journal of Labor Economics 33(3), 683–710.
27
Staiger, D. and Stock, J. H.: 1997, Instrumental variables regression with weak instruments, Econo-
metrica 65(3), 557–586.
Thakral, N. and To, L. T.: 2017, Daily labor supply and adaptive reference points, Working paper.
Ziliak, J. P. and Kniesner, T. J.: 2005, The effect of income taxation on consumption and labor
supply, Journal of Labor Economics 23(4), 769–796.
28
Table 1: Descriptive Statistics.
Panel A - Entire SampleMean SD Min Max Count
Distance 229.2052 73.07427 0 343.7 829Percentage Aged 18 and 64 59.39107 8.066361 19.2 95.8 829Income 7538.242 1322.182 4182.561 11086.62 829Men’s Weekly hours worked 46.45911 4.876871 22.4 67.4 829Women’s Weekly hours worked 35.10084 4.344163 7.3 51.8 829Men’s Labor Force Participation Rates 73.86719 11.5267 31.1 100 829Women’s Labor Force Participation Rates 68.27382 15.32451 1.3 100 829Percentage who Hold an Academic Degree 25.97479 13.4897 1.5 68.9 829Panel B - 320 km
Mean SD Min Max CountDistance 208.1978 60.96126 0 317.9 692Percentage Aged 18 and 64 58.78801 7.882589 19.2 95.8 692Income 7760.304 1320.407 4182.561 11086.62 692Men’s Weekly hours worked 46.47168 4.789316 22.4 66.7 692Women’s Weekly hours worked 35.0854 4.309348 7.3 51.8 692Men’s Labor Force Participation Rates 74.0091 11.44104 31.1 100 692Women’s Labor Force Participation Rates 68.32645 15.23294 1.3 97.4 692Percentage who Hold an Academic Degree 26.83223 13.78233 1.5 68.9 692Panel C - 300 km
Mean SD Min Max CountDistance 203.5718 57.82338 0 297.6 664Percentage Aged 18 and 64 58.78283 7.946593 19.2 95.8 664Income 7696.2 1309.731 4182.561 11086.62 664Men’s Weekly hours worked 46.45527 4.841269 22.4 66.7 664Women’s Weekly hours worked 35.09066 4.335908 7.3 51.8 664Men’s Labor Force Participation Rates 74.04849 11.43837 31.1 100 664Women’s Labor Force Participation Rates 68.31943 14.94758 6.8 97.4 664Percentage who Hold an Academic Degree 26.23072 13.26946 2 65.9 664Panel D - 280 km
Mean SD Min Max CountDistance 188.9071 49.40015 0 279 567Percentage Aged 18 and 64 58.51658 7.923274 19.2 86.4 567Income 7646.755 1305.175 4881.104 11086.62 567Men’s Weekly hours worked 46.52205 4.901982 22.4 66.7 567Women’s Weekly hours worked 35.07284 4.417525 7.3 51.8 567Men’s Labor Force Participation Rates 74.15097 11.39835 31.1 100 567Women’s Labor Force Participation Rates 68.17196 14.87967 7.2 97.4 567Percentage who Hold an Academic Degree 26.0843 13.49039 2.1 65.9 567
Notes: This table presents descriptive statistics for the variables used in the empirical analysis.
29
Table 2: First Stage Results.
Dependent Variable - Log IncomePanel A: Entire Sample
(1) (2)Without Controls With Controls
Instrument 0.110*** 0.186***(0.0198) (0.0405)
Observations 832 810R-square 0.0224 0.386
First Stage F- statistic 30.88 21.06
Panel B: 320 kmWithout Controls With Controls
Instrument 0.272*** 0.329***(0.0228) (0.0425)
Observations 695 676R-square 0.0448 0.416
First Stage F- statistic 141.69 59.76
Panel C: 300 kmWithout Controls With Controls
Instrument 0.233*** 0.300***(0.0254) (0.0444)
Observations 667 651R-square 0.0221 0.415
First Stage F- statistic 84.47 45.72
Panel D: 280 kmWithout Controls With Controls
Instrument 0.242*** 0.309***(0.0311) (0.0483)
Observations 567 557R-square 0.0241 0.438
First Stage F- statistic 60.77 40.79
Notes: This table presents the First Stage results. All the regressions include the distance in kilo-meters from the HaArava district. The other controls variables are: the proportion of people aged18 to 64; the proportion of people who hold academic degree; and year of establishment. Robuststandard errors are in parentheses. * indicates significance at the 10% level; ** indicates signifi-cance at the 5% level; and * indicates significance at the 1% level.
30
Table 3: OLS versus 2SLS Results.
Dependent Variable: Men’s Log Weekly hours worked
Panel A: EntireSample
Panel B: 320 km Panel C: 300 km Panel D: 280 km
(1) (2) (3) (4) (5) (6) (7) (8)OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS
Log Income 0.0696* 0.645** 0.0551 0.316** 0.0496 0.235* 0.0306 0.464***(0.0366) (0.306) (0.0431) (0.126) (0.0443) (0.132) (0.0325) (0.161)
Observations 810 810 676 676 651 651 557 557
Notes: This table presents OLS and 2SLS results for each sample. All the regressions include the distancein kilometers from the HaArava district and other control variables, which are: the proportion of peopleaged 18 to 64; the proportion of people who hold academic degree; and year of establishment. Robuststandard errors are in parentheses. * indicates significance at the 10% level; ** indicates significance atthe 5% level; and * indicates significance at the 1% level.
Table 4: OLS versus 2SLS Results: Sub-Samples Divided by First Quartiles of Wealth andPercentage of Elderly People; and Including Squared Distance.
Dependent Variable: Men’s Log Weekly hours worked
Panel A:Percentageof PeopleAged 18-64Above theFirstQuartile
Panel B:Percentageof PeopleAged 18-64Below theFirstQuartile
Panel C:Non-WealthyMunicipali-ties
Panel D:Wealthy Mu-nicipalities
Panel E:IncludingDistanceSquared
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS
Log Income 0.089** 0.59* -0.35 -0.27 -0.25 2.83* 0.1** 0.31 0.0745* 0.242***(0.043) (0.31) (0.21) (0.2) (0.29) (1.53) (0.04) (0.25) (0.0397) (0.0788)
Observations 585 585 223 223 172 172 638 638 810 810
Notes: This table presents OLS and 2SLS results from using as cutoffs the first quartile values of wealthand percentage of elderly people (Panel A-D), as well as results from inclusion of squared distancefrom the HaArava district (Panel E). All the regressions include the distance in kilometers from theHaArava district and other control variables, which are: the proportion of people aged 18 to 64; theproportion of people who hold academic degree; and year of establishment. Robust standard errors arein parentheses. * indicates significance at the 10% level; ** indicates significance at the 5% level; and *indicates significance at the 1% level.
31
Table 5: Difference-in-Differences.
Panel A: Differences between MeansHaArava Outside
HaAravaDiff. Std. Error
The Proportion of Men who Hold AnAcademic Degree
19.42 (3.37) 22.66 (0.45) 3.23 (5.05)[7] [841]
The Proportion of women who HoldAn Academic Degree
29.2 (3.37) 29.56 (0.51) 0.36 (5.72)[7] [851]
Men Median Age 29.71 (1.56) 28.41 (0.22) -1.29 (2.5)[7] [886]
Women Median Age 28.71 (1.79) 29.79 (0.24) 1.08 (2.79)[7] [886]
The Proportion of Men Aged 18 to 64 63.02 (2.94) 59.35 (0.28) -3.67 (2.23)[7] [886]
The Proportion of Women Aged 18 to 64 62.78 (2.86) 59.35 (0.28) -3.43 (3.2)[7] [886]
The Proportion of Men who areMarried
60.68 (2.38) 56.52 (0.33) -4.15 (3.75)[7] [885]
The Proportion of Women who areMarried
63.17 (2.62) 57.46 (0.34) -5.7 (3.83)[7] [885]
Panel B: hours workedEntire Sample
withoutControls
Entire Samplewith Controls
320 km 300 km 280 km
Difference-in-Differences Coefficient 5.478** 5.537* 5.531* 5.548* 5.473*(2.644) (3.018) (3.063) (3.080) (3.129)
Observations 1677 1632 1352 1302 1114R-square 0.601 0.757 0.775 0.778 0.797
Panel C: Labor Force Participation RatesEntire Sample
withoutControls
Entire Samplewith Controls
320 km 300 km 280 km
Difference-in-Differences Coefficient -3.185 -2.275 -2.220 -2.427 -2.711(3.236) (3.687) (3.747) (3.769) (3.834)
Observations 1678 1632 1352 1302 1114R-square 0.0458 0.529 0.558 0.566 0.600
Notes: Panel A presents t-test results for differences between means. Standard errors are in parentheses;numbers of observations are in brackets. Panels B and C present the difference-in-differences results ofEquation (3) in the text. All the regressions include the following control variables: the proportion ofpeople aged 18 to 64; the proportion of people who hold academic degree; and year of establishment.Robust standard errors are clustered at the locality level and are in parentheses. * indicates significanceat the 10% level; ** indicates significance at the 5% level; and * indicates significance at the 1% level.
32
Table 6: OLS versus 2SLS Results: Men’s Log Labor Force Participation Rates.
Panel A: EntireSample
Panel B: 320 km Panel C: 300 km Panel D: 280 km
(1) (2) (3) (4) (5) (6) (7) (8)OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS
Log Income -0.00937 0.518** 0.0304 0.285** 0.0149 0.145 0.0421 0.144(0.0565) (0.255) (0.0622) (0.118) (0.0649) (0.124) (0.0734) (0.119)
Observations 810 810 676 676 651 651 557 557Notes: This table presents OLS and 2SLS results for men’s log labor participation rates. All the regres-sions include the distance in kilometers from the HaArava district and other control variables, whichare: the proportion of people aged 18 to 64; the proportion of people who hold academic degree; andyear of establishment. Robust standard errors are in parentheses. * indicates significance at the 10%level; ** indicates significance at the 5% level; and * indicates significance at the 1% level.
33
Table 7: The Effect of the Sabbath Year on Women’s Labor Supply: 2SLS Estimates .
Women’s LogWeekly hours
worked
Women’s Log LaborForce Participation
RatesEntire SampleLog Income -0.22 0.13
(0.4) (0.32)Observations 810 810
320kmLog Income -0.22 0.18
(0.19) (0.19)Observations 676 676
300kmLog Income -0.31 0.17
(0.21) (0.22)Observations 651 651
280kmLog Income -0.25 0.06
(0.17) (0.19)Observations 557 557
Notes: This table presents the effect of the Sabbath year on women’s hours worked and labor par-ticipation rates. All the regressions include the distance in kilometers from the HaArava districtand other control variables, which are: the proportion of people aged 18 to 64; the proportionof people who hold academic degree; and year of establishment. Robust standard errors are inparentheses. * indicates significance at the 10% level; ** indicates significance at the 5% level; and* indicates significance at the 1% level.
34
Table 8: The Effect of the Sabbath Year on Alternative Outcomes: Placebo Tests.
(1) (2) (3)The Proportion ofPeople Who Lived
Within the Locality FiveYears Age
The Proportion ofPeople Who Lived
Outside the LocalityFive Years Ago
The Percentage whoHold an Academic
Degree
Entire Sample
Indicator for beingOutside Biblical
Israel
2.481 -3.821 -4.173
(11.47) (11.49) (3.337)
Observations 829 820 830
R-square 0.0192 0.0217 0.00289320 km
Indicator for beingOutside Biblical
Israel
4.268 -6.307 0.307
(11.52) (11.56) (3.605)
Observations 692 688 693
R-square 0.0243 0.0302 0.00391300 km
Indicator for beingOutside Biblical
Israel
4.641 -6.746 -3.405
(11.55) (11.59) (3.571)Observations 664 660 665
R-square 0.0240 0.0296 0.000926280 km
Indicator for beingOutside Biblical
Israel
5.812 -8.581 -5.080
(11.64) (11.71) (3.847)
Observations 564 562 567
R-square 0.0222 0.0293 0.00314Notes: This table presents the effect of being outside Biblical Israel on different outcomes for differentdistances (entire sample, within 320 km, 300 km, and 280 km) from the HaArava district. All the re-gressions include the distance in kilometers from the HaArava district. Robust standard errors are inparentheses. * indicates significance at the 10% level; ** indicates significance at the 5% level; and *indicates significance at the 1% level.
35
Table 9: Pre- and Post-Treatment Differences in Incomes and Difference-in-DifferencesEstimation for Income.
Mean(HaArava)
Mean (OutsideHaArava)
Difference Std. Error Obs.
Panel A: Real IncomeAfter 2008 (2009-2013)
9114.1869 8437.8409 -676.3461 798.2727 268(236.56) (109.791)
[5] [263]
Panel B: Real IncomeBefore 2008 (2002-2007)
8190.6281 7491.1239 -699.5042 733.5670 322(681.53) (100.231)
[6] [316]
Panel C: Differences-in-Differences Estimates Using the Years2002-2013 (Dependent Variable - Income in 2011 Terms)
(1) (2) (3) (4)Entire Sample 320 km 300 km 280 km
Difference-in-Differences
973.2*** 958.7*** 952.6*** 932.3***
(336.6) (336.2) (335.8) (335.4)
Observations 624 554 542 470
R-square 0.954 0.954 0.954 0.948Notes: Panels A and B of this table present pre- and post-treatment differences between income meansusing the years 2002-2013. Standard errors are in parentheses; numbers of observations are in brack-ets. Panel C shows Difference-in-Differences estimation results for income. Throughout estimations weadded year and distance dummies. Incomes are in 2011 terms according to the following formula: In-come in 2011 terms = Income in current year (CPI in 2011)/(CPI in current year). * indicates significanceat the 10% level; ** indicates significance at the 5% level; and * indicates significance at the 1% level.
36
Figure 1: Mapping of Israel’s Districts.
Notes: This figure presents a mapping of contemporary Israel based on its main districts. TheHaArava district (in dark), which is below the Dead Sea, is defined as the area outside the Biblicalland of Israel. All other districts are defined as the area inside the Biblical land. Source: Googleimages labeled for reuse with modification.
37
Figure 2: Average Income in 2008: HaArava District versus Other Districts.
02,0
00
4,0
00
6,0
00
8,0
00
Avera
ge Incom
e in N
IS
HaArava District All Other Districts Districts within 280km Districts within 300km Districts within 320km
Notes: This figure presents incomes for localities within different distances (entire sample, within320 km, 300 km, and 280 km) from the HaArava district and for localities within the HaAravadistrict for 2008.
38
Figure 3: The Percentage who Hold an Academic Degree: Treated versus ControlGroups.
010
20
30
The P
erc
enta
ge w
ho H
old
an A
cadem
ic D
egre
e
HaArava District All Other Districts Districts within 280km Districts within 300km Districts within 320km
Notes: This figure presents population percentages holding an academic degree in localitieswithin different distances (entire sample, within 320 km, 300 km, and 280 km) from the HaAr-ava district and for localities within the HaArava district for 2008.
39
Figure 4: Average Real Income by District and Year.
02,0
00
4,0
00
6,0
00
8,0
00
10,0
00
Avera
ge Incom
e in R
eal T
erm
s
Before 2008 During 2008 After 2008
All O
the
r D
istr
icts
Ha
Ara
va
Dis
tric
t
All O
the
r D
istr
icts
Ha
Ara
va
Dis
tric
t
All O
the
r D
istr
icts
Ha
Ara
va
Dis
tric
t
Notes: This figure presents income differences between localities inside and outside theHaArava district before, during, and after the 2008 Sabbath year.
40